NVIDIA Releases Alpamayo 2 Super: Open Reasoning Model for Autonomous Driving with Commercial License

Summary:NVIDIA has officially released Alpamayo 2 Super, the most powerful open reasoning model in its Alpamayo family, designed for autonomous driving and robotaxis. Built on NVIDIA Cosmos 3 Super Reasoner and post-trained with reinforcement learning, it supports full-surround camera input and outputs five key components including trajectory, chain-of-causation, and meta-actions. Alpamayo 2 Super ranks first on the LingoQA benchmark, significantly outperforming Qwen2.5-VL 72B, Gemini 2.5 Pro, and GPT-4

---

Key Highlights

• Open Commercial License: Available under the Linux Foundation's OpenMDW-1.1 license, allowing fine-tuning, derivative models, and commercial redistribution, giving AV developers full control.
• Benchmark-Leading Performance: Ranks first on the LingoQA autonomous driving reasoning benchmark, outperforming Qwen2.5-VL 72B by 17.0 points, Gemini 2.5 Pro by 15.1 points, and GPT-4o by 23.2 points.
• Multitask Capabilities: A single model supports trajectory planning, chain-of-causation reasoning, meta-action prediction, auto-labeling, and visual question answering, simplifying the development stack.
• 360-Degree Perception: Fuses front, side, and rear camera views for enhanced understanding of complex scenarios.

Technical Architecture and Innovation

Alpamayo 2 Super is built on NVIDIA Cosmos 3 Super Reasoner and post-trained with reinforcement learning. The model is 3x the scale of its predecessors (approximately 30 billion parameters), significantly improving generalization from sparse examples, which is critical for handling long-tail events in autonomous driving.

The model processes full-surround camera input, fusing multiple views to achieve deep understanding of lane changes, merges, unprotected turns, and complex intersections.

Multitask Outputs and Safety Validation

For each driving scenario, Alpamayo 2 Super generates five tightly coupled outputs:

1. Trajectory: The vehicle's planned path.
2. Chain-of-Causation (CoC): Explains the reasoning behind the decision.
3. Meta-action: Intent such as yield, lane change, stop.
4. Reasoning auto-labels: Generates CoC annotations for training and validation data.
5. Visual question answering: With 2D visual grounding, linking answers to specific image regions.

These outputs enable developers to understand the model's decision-making process and integrate with NVIDIA Halos safety-validation workflows, aligning with ISO/PAS 8800 requirements for robust AV safety engineering.

Industry Impact and Developer Recommendations

The release of Alpamayo 2 Super marks a significant step toward open, customizable AI models for autonomous driving. Developers can leverage the model in cloud-based workflows to generate high-quality reasoning traces and synthetic data for distillation, then deploy efficient specialized models in vehicles, enabling a complete cloud-to-car workflow.

NVIDIA also provides complementary tools such as AlpaSim (closed-loop simulation), AlpaGym (reinforcement learning), and Physical AI Open Datasets to accelerate model development and validation.

Conclusion

Alpamayo 2 Super is a key part of NVIDIA's strategy in autonomous driving AI. Its open commercial license and leading performance will drive innovation, lower development barriers, and accelerate the deployment of safe and reliable autonomous driving technology.

Based on NVIDIA official blog report.

---

Source: NVIDIA AI News (2026-08-04)

— END —

Tags: #NVIDIA #Autopilot #Open Source Model #Logic Model #Alpamayo